Porting Moebius 0.2B Image Restoration Model to Browser with Claude Code
Decision Brief
What changedAuthor uses Claude Code to port the Moebius 0.2B image restoration model, originally dependent on PyTorch and CUDA, to run in browsers with WebGPU support.
Why it mattersAI builders can learn how to deploy small models to browsers using a coding agent, with hands-on experience in ONNX conversion and WebGPU integration.
Who should careAI coding tool users, Inference / infra teams
Affected stackClaude CodeHugging FaceNVIDIA
Builder actionEvaluate
Source confidenceMedium · Reliable media or first-hand reporting
After spotting the lightweight Moebius 0.2B image restoration model on Hacker News, the author decided to port it to the browser. Claude analyzed feasibility and suggested using ONNX Runtime Web with WebGPU. The author prepared source code and weights, then launched Claude Code for conversion and web development. Claude Code automatically converted the PyTorch model to ONNX, uploaded it to Hugging Face, and built the frontend UI. The author only tested and gave feedback, writing zero code. The result is live at simonw.github.io/moebius-web/, supporting Chrome, Firefox, and Safari.
Summary basis: official / RSS sourceUnless it says 'full article read', this summary is based only on publicly available content — it never pretends to have read restricted originals.
Sources
- Simon Willison：Blog
Hands-on notes on LLM tools, local models, and practical AI engineering.
- Simon Willison：Blog